KetchUp is a free, open-source 3D modeling app for woodworking and furniture design, written in Rust and rendered with WebGPU, running on Windows, macOS, Linux and desktop browsers with no account required. It targets precise dimensioned planning of shelves, tables and cabinets as a lightweight SketchUp alternative, though textures, scenes and some model features are unsupported and OBJ export omits materials.
WebGPU
A modern API for high-performance computing and graphics rendering in browsers
Curated Intelligence
Google released the open-source multimodal embedding model EmbeddingGemma 2, which outperforms rivals twice its size on benchmarks despite having only 740 million parameters. The model supports local execution with minimal RAM usage (191 MB), significantly reducing vector database storage requirements and improving retrieval efficiency.
The creator of React Bits has released Canvas UI, the first open-source component library built on the experimental HTML-in-Canvas API. It offers 35 GPU-accelerated effects (fluids, fire, etc.) with dual WebGL/WebGPU backends that capture live DOM as textures while preserving accessibility and interactivity. Distributed via a shadcn-compatible registry, it also supports MCP for AI agent integration.
Hugging Face released @huggingface/kernels and a collection of 207 optimized kernels to accelerate local AI inference in browsers. The tool simplifies WebGPU integration for model execution by providing low-level GPU operation support. Developers can directly reuse these kernels to improve inference speed and compatibility in web applications.
Hugging Face officially released Transformers.js v4, featuring a completely rewritten C++-based WebGPU runtime. Developed in close collaboration with the ONNX Runtime team, this update significantly improves performance, accuracy, and coverage for approximately 200 supported model architectures, marking a major leap in browser-side inference capabilities.
Hugging Face has released Transformers.js v3, featuring native WebGPU support that delivers up to 100x faster inference compared to WASM. The update also introduces new quantization formats, supports 120 model architectures, and ensures compatibility with Node.js, Deno, and Bun, significantly lowering the barrier for deploying large models in web environments.